Formulas and the Building Blocks of Ṭhumrī Style—A Study in “Improvised” Music

Chloe Zadeh

PART I: IMPROVISATION, STRUCTURE AND FORMULAS IN NORTH INDIAN CLASSICAL MUSIC

usicians, musicologists and teenage backpackers alike speak of “improvisation” in M modern-day Indian classical music.1 Musicians use the term to draw attention to differences between different portions of a performance, distinguishing between the

“composition” of a particular performance (in vocal music, a short melody, setting a couple of lines of lyrics) and the rest of the performance, which involves musical material that is different in each performance and that musicians often claim to have made up on the spur of the moment.2 In their article “Improvisation in Iranian and Indian music,” Richard Widdess and Laudan Nooshin draw attention to modern-day Indian terminology which reflects a conceptual distinction between improvised and pre-composed sections of a performance

(2006, 2–4). The concept of improvisation has also been central to Western imaginings of

North Indian classical music in the twentieth century. John Napier has discussed the long history of Western descriptions of Indian music as “improvised” or “extemporized,” which he dates back to the work of Fox Strangways in 1914. He points out the importance the concept took on, for example, in the way in which Indian music was introduced to mainstream

Western audiences in the 1960s; he notes that the belief that the performance of Indian music

1 I am very grateful to Richard Widdess and Peter Manuel for their comments on draft versions of this article. 2 The relationship between those parts of the performance that musicians label the “composition” and the rest of the performance is actually slightly more complicated than this; throughout, a performance of North Indian classical music is normally punctuated by a refrain, or mukhra, which is a part of the composition. Even apparently improvised parts of the performance may also make extensive use of material that is derived from the composition. Formulas and the Building Blocks of Ṭhumrī Style constituted an improvised, meditative act allowed Indian music to become “a metaphor for a particular kind of psychic or spiritual freedom” (2006, 3).

The term “improvisation” evokes ideas of spontaneity and freedom; it would seem antithetical to the idea of a carefully worked-out and pre-planned structure. These connotations, however, are potentially misleading when the term is used in an Indian classical context.3 Contrasting with the idea of whimsical freedom that the word “improvisation” evokes, a number of scholars have drawn attention to ways in which performances in the

North Indian classical tradition are highly structured. John Napier, for example, is highly critical of some of the “misunderstandings” that arise from Western audiences’ impressions of North Indian classical music as improvised. He is suspicious of what he calls a “crude racialization,” which, he feels, leads some to view Indian music as “structurally deficient.” In contrast with this view, he has identified what he calls a process of “intensification” in North

Indian classical music; this structures performances such that they normally start slowly and in a low register, becoming increasingly fast, complex and exploring ever higher pitches as the performance progresses (2006, 5). Similarly, in their discussion of improvisation in

Indian music, Widdess and Nooshin have drawn attention to “fundamental processes of development” and “compositional principles” that structure North Indian classical music

(2006, 7–8). One of these is vistār, a structural principle which occurs throughout North

Indian classical music. Vistār is one part of Napier’s “intensification” and is the process in

3 It is not only in studies of North Indian classical music that scholars have pointed to the problems inherent in the use of the concept of “improvisation” to make sense of different types of music. Bruno Nettl (1974) has critiqued the conventional opposition between composition and improvisation, suggesting that they would better be modeled as opposite poles of a spectrum, on which most musical traditions sit somewhere in the middle, involving a combination of pre-planned and spontaneous musical events. Moreover, Laudan Nooshin (2003) has discussed the political implications of the discursive distinction between improvisation and composition, suggesting ways in which it is informed by a Western, orientalist ideology.

2 Analytical Approaches To World Music 2.1 (2012) which musicians gradually introduce the rāg of the performance to their audiences, by focusing on successively higher (and sometimes lower) pitches in turn (2006, 7).4

Elsewhere, Widdess draws upon theories of cognitive schemas in order to make sense of how some of these “compositional principles” operate in North Indian classical music. A schema is a memory structure. In a discussion of schemas and music in his book Music and

Memory, Bob Snyder defines it as follows: “When a number of different situations occurring at different times seem to have aspects in common, they are eventually averaged together into an abstract memory framework.... Built up out of the commonalities shared by different experiences, these frameworks are referred to as ‘schemas’” (2000, 95). Cognitive psychologists suggest that schemas operate widely in our everyday lives, acting as scripts to generate typical patterns of behavior in familiar situations. Noting that schema theory has been useful in accounting for features observed in the analysis of various types of music,

Widdess discusses schemas in North Indian classical music in particular. He focuses his analysis on one ālāp performance by the sitarist Budhaditya Mukherjee. There, he identifies what he calls “pitch schemas,” which he defines as “the static, quasi-spatial, hierarchical relationships among a group of defined pitches (such as a scale)” and “contour schemas,” which he defines as “a temporal sequence of pitches underlying, and repeatedly embellished or varied in, a group of melodic phrases” (2011, 194). He also notes that the process of vistār itself is a schema (206–07).

In this article, I focus on the North Indian semi-classical genre, ṭhumrī. As a “semi- classical” genre (as opposed to the fully classical genres khyāl and dhrupad), ṭhumrī is not bound by the strict rules concerning rāg that apply to its classical counterparts. Rather,

4 When I was learning to sing Indian classical music, this overall pattern would inform not only the pieces my teacher taught me, but even the way she structured my lessons. We would start with long, slow exercises focusing on middle Sa, then explore lower register, and then start a series of exercises which reached ever higher notes while increasing in speed and complexity. By learning in this way, this overall structural progression came to seem perfectly natural to me.

3 Formulas and the Building Blocks of Ṭhumrī Style musicians often talk about the flexibility and freedom permitted in the genre. Nevertheless, many of the schemas and “compositional principles” that Widdess identifies in classical genres also operate in the semi-classical ṭhumrī. Vistār, for example, informs the development of most ṭhumrī performances, though it occurs in a less explicit way than in

ālāps such as the one that Widdess analyzes. In ṭhumrī, vistār is usually telescoped so as to occur relatively quickly, it rarely involves any exploration of the register below madhya Sa

(the middle tonic) and it sometimes does not structure the whole performance, instead only featuring in certain sections. In addition, ṭhumrī performances are structured by a large-scale formal schema, which dictates the order of musical events in a ṭhumrī performance. Figure 1 is a summary of the structural units that make up ṭhumrī’s overall form, showing the different structural options available to performers. In a performance of ṭhumrī, the performer might follow the arrows along any possible course. This diagram indicates the flexibility inherent in schematic organization: certain events may be missed out, while others may be repeated, extended or shortened. In addition to these two large-scale schemas, other pre-existing constraints on performances of ṭhumrī include the composition, which is a pre-existing melody, the typical style of ornamentation expected in the genre and the rāg and tāl of the performance.

It is not only in their use of relatively fixed compositions and in the presence of large- scale structural principles that North Indian classical performances display evidence of advance planning. In addition to this, musicologists have also drawn attention to the large amount of memorized and rehearsed musical material that North Indian classical musicians use in apparently improvised portions of their performances. Stephen Slawek, for instance, has described improvisation in North Indian classical music as a combination of memorized patterns, material based on pre-existing models and material produced by learned, generative

“programs” (1998, 363). Widdess and Nooshin, too, have drawn attention to the frequent use

4 Analytical Approaches To World Music 2.1 (2012)

Figure 1. The structure of a ṭhumrī performance.

Glossary of terms

Ālāp: An unmetered introductory passage, concluding when the tabla enters. Sthāyī: The first line of the composition, consisting of a set of lyrics and an approximate melodic outline. Bol banāo: Passages which use the words of the composition in various melodic settings, designed to reveal the emotional nuances inherent in each phrase. Antarā: A subsequent line of the composition, whose melody often explores the

register surrounding the upper tonic. Laggī: A faster, virtuosic passage in which the tabla often takes on a solo role.

of memorized musical material in Indian classical performances; they suggest that although musicians may not always create entirely new material on the spur of the moment throughout the performance, nevertheless spontaneity lies in “the decision to use a particular phrase or sequence of phrase at a given moment in the performance” (2006, 3). In ṭhumrī in particular,

Peter Manuel has highlighted musicians’ repeated use of “favorite” musical phrases in different performances of ṭhumrīs in the same rāg (1989, 135–36). Napier (2006, 1), Wim van der Meer (1980, 143), Daniel Neuman (1980, 23) and Thom Lipiczky (1985) are amongst the variety of other scholars who have also highlighted this phenomenon in North

Indian classical music.

In Slawek’s view, using memorized material offers musicians a way of meeting the requirement that they “keep it going” in performance: that is, that they continue to generate musical material at length, over the course of a performance that might last for some hours.

Manuel makes a similar point about singers’ use of memorized phrases in performances of

5 Formulas and the Building Blocks of Ṭhumrī Style

ṭhumrī in particular, writing that “it is essential that the vocalist has sufficient mastery and facility with these phrases, that they be at his fingertips, as it were, so that when he is actually concentrating on the musical sentiment of the text for inspiration, they emerge naturally, spontaneously, and with continuity and feeling” (1989, 140). Slawek draws a link between the use of memorized patterns in North Indian classical music and the means by which it is customarily learned and practiced. He writes, “much of what occurs in a performance of

Hindustānī music … is actually ‘fixed’ music (one might alternately say memorized) in the sense that the performer has practiced and rehearsed those exact melodic or rhythmic phrases hundreds, if not thousands, of times before” (1998, 336). Widdess and Nooshin, too, discuss the relationship between improvisation in North Indian classical music and the way in which it is learned: they note, for example, that the memorization of relatively fixed musical material frequently plays a large role in being taught to perform North Indian classical music

(2006, 109). My own experiences learning to sing North Indian classical music with the vocalist Sunanda Sharma confirm this.

My analysis has revealed a very large amount of formulaic musical material in ṭhumrī performances in the twentieth century. In this essay, I will explore its significance from a music-analytical perspective. I will consider possible relationships between formulas in

ṭhumrī and those that appear in oral literature, everyday speech and other musical traditions and consider how theories developed in relation to formulas in these other domains may shed light on formulas in ṭhumrī. Based on my analysis of a large number of ṭhumrī recordings, I will then discuss the particular types of formulas that occur in ṭhumrī and examine how they operate in performance. I will argue that formulas are the building blocks of ṭhumrī and suggest that identifying the different types of formulas that occur in the genre offers a useful starting point for its analysis.

6 Analytical Approaches To World Music 2.1 (2012)

PART II: FORMULAS IN MUSIC, ORAL LITERATURE AND EVERYDAY SPEECH

Formulas in Improvised Music

North Indian classical music is not the only seemingly improvised musical tradition in which musicians repeatedly make use of memorized musical formulas. Their presence is well-documented in improvisation in jazz, as is the practice of memorizing solos as a way of learning how to improvise. In his book Thinking in Jazz, for example, Paul Berliner discusses the musical patterns that jazz musicians employ repeatedly in their improvisations. He writes that, “many students begin acquiring an expansive collection of improvisational building blocks by extracting those shapes they perceive as discrete components from the larger solos they have already mastered and practicing them as independent figures” and that “they acquire others selectively by studying numerous performances of their idols” (1994, 101).

Like Slawek in his description of North Indian classical music, Berliner highlights the role that memorized material plays in enabling musicians to generate solos on the spur of the moment: he notes that veteran improvisers will possess a “storehouse” or “vocabulary” of musical ideas, which “[provides] readily accessible material that meets the demands of composing music in performance.” He cites one of his informants, who laughed about the necessity of being able to turn to memorized material “when your mind is crippled and you can’t think of anything to play” (1994, 102).

In a cross-cultural study of musical improvisation, Aaron Berkowitz draws attention to the occurrence of musical formulas in a variety of musical traditions. Like Widdess and

Nooshin, Slawek and Berliner, Berkowitz also links the presence of formulas in improvised music with the large amount of memorization that is often involved in learning how to improvise. He characterizes formulas as an aid to the performer who needs to produce complex musical performances on the spur of the moment. He writes that these musical

7 Formulas and the Building Blocks of Ṭhumrī Style patterns, once “internalized” and “automatized,” should be able to be “performed instantly and without preplanning” (2010, 42), giving the impression that the performer has “brains in the fingers” (2010, 37–38).

Formulas in Oral Literature

The first and most well known discussions of formulas and their relationship with improvisation occurred in the field of linguistics and were concerned specifically with oral poetry. Early in the twentieth century, Milman Parry drew attention to formulas in Homeric epic poetry; he suggested that their presence indicated that this poetry originated in an oral poetic tradition (see his collected papers in Parry (1987)). His work, along with that of his student Albert Lord on still-performed Slavic oral epic poetry (1965), has been highly influential on scholars working subsequently on a variety of orally transmitted genres. Walter

Ong, for example, draws on Parry’s hypothesis in order to make a cross-cultural argument about the impact of the advent of writing on the development of societies over time. Writing of the differences between “orality” and “literacy,” he highlights formulas as an example of one of the characteristics of oral traditions, contrasting them with written literature (1982).

More recently, the cognitive psychologist David Rubin has conducted an in-depth study of the relationship between memory and the characteristic structural features of performances in oral traditions (1995). As is Widdess’ analysis of North Indian classical music, Rubin draws attention to the role of cognitive schemas in structuring performances: he suggests that it is through schemas that performers memorize and then generate performances in oral traditions. Rubin employs a variety of technical terms, some borrowed from other scholars, in order to distinguish between different types of schemas. One such type is a “script,” a term which he takes from the work of Schank and Ableson (1977) and which he defines as “a

8 Analytical Approaches To World Music 2.1 (2012) predetermined, stereotyped sequence of actions that defines a well-known situation” (1995,

24). He draws attention to the presence of scripts in a number of oral traditions.

Drawing heavily on the work of Parry and Lord, Rubin also suggests that schema theory might account for the widespread presence of formulas in oral literature. Specifically, he argues that formulas are a result of the multiple, combining constraints that schemas produce in performance. In the case of oral poetry, for example, Rubin notes that large-scale structural schemas (such as the overall plot of the poem) constrain the performer, limiting the directions the performance may take. He also draws attention to other possible constraints in performance, including meter, rhyme and the conventional vocabulary from which the poet may draw. He points out that these multiple, simultaneous constraints severely limit the material that a performer might use. Formulas are tailor-made phrases which obey these constraints. They might consist, for example, of conventional, multi-word epithets, used in the description of particular characters, which also happen to fit appropriately in a particular part of a metrical cycle. The poet has a large repertoire of these formulas to hand and uses them as appropriate. The poet does not, therefore, have to create material that obeys all the necessary constraints entirely from scratch during the performance.

According to Rubin’s theory, schemas can operate at different structural levels. As well as large-scale, global schemas which structure whole performances and medium-scale

“scripts,” which dictate the order of a well-known known sequence of events, for example the events surrounding a hero’s preparing his army for battle (1995, 24–28), Rubin also highlights the operation of schemas at a much smaller scale within performances, drawing attention to what he calls “surface schemas” (70–72). It is possible to think of formulas as an example of this more local type of patterning; according to this model, formulas are not only a result of the operation of schemas in performance (and the constraints they produce), but are also schemas themselves.

9 Formulas and the Building Blocks of Ṭhumrī Style

A number of scholars have drawn a connection between the performance of oral poetic traditions and improvisation in music. Leo Treitler, for example, suggests taking the theory of

Parry and Lord about oral composition “as a paradigm” for the study of Gregorian plainchant, looking at ways in which notated plainchants (like notated Homeric epics) display evidence of a past tradition of oral transmission (1974). Like Rubin, he also draws upon cognitive psychology in order to support the view that formulas appear in the products of oral traditions as a result of the characteristics of human memory. Others have applied the work of Parry and Lord to the study of jazz improvisation. (See, for example, Smith 1983.) Berkowitz draws a general, cross-cultural comparison between improvised music and the performance of oral epic poetry, drawing particular attention to the role that formulas play in the transmission of non-notated traditions (2010, 27–28).5

North Indian classical music is, for the most part, transmitted orally. Its repertoire is stored in the memories of the musicians who perform it, not in written scores. Every piece is generated anew in each performance, such that the exact order of pitches and note-lengths used will never be the same in any two performances, even if they are of the same piece.

Nevertheless, performances do conform to a number of rules and schemas, such as those associated with the rāg and tāl of the composition. Like performers of oral poetry, North

Indian classical musicians are required to produce seemingly novel material at speed and on the spur of the moment, which nevertheless also obeys a variety of different structural and

5 Richard Widdess (2011 personal communication) has pointed out a difference of emphasis in some of these different studies of orality and music: while Treitler and Rubin are interested primarily in how people remember largely fixed oral texts, either verbal or musical, other scholars, including Parry/Lord and Berliner, focus instead on how people improvise partially new texts in performance. This is partly a result of the different subject matter they study. Treitler, for example, examines Gregorian chant, the melodies of which differ only very slightly between different written versions, implying that they must have been transmitted by almost total memorization. Berliner, meanwhile, studies jazz improvisation, a tradition in which great value is attached to novelty and originality in performance. In these different studies, formulas and/or schemas would seem to play two different, albeit related, roles: first, to aid the recall of fixed texts and, second, to provide the means by which performers can generate long, seemingly improvised, texts on the spur of the moment. In both cases, the memorization of formulaic chunks of material would seem to be crucial in facilitating successful performances.

10 Analytical Approaches To World Music 2.1 (2012) other constraints. Just as formulas in oral poetry would seem to provide a solution to this problem, so the formulaic material that occurs in North Indian classical music would seem to offer a comparable solution to a comparable problem.

Formulas in Everyday Speech

Recent work in the field of linguistics has revealed ways in which everyday spoken language contains formulas comparable to those in oral literature. Most of this work is indebted to Andrew Pawley and Frances Syder’s 1983 book chapter “Two Puzzles for

Linguistic Theory.” They highlight the existence in conversational speech of what they call

“sentence stems,” which are a type of formulaic linguistic unit. They argue that the knowledge of a body of such units is necessary for “the fluent and idiomatic control of a language” (191). They propose that paying attention to sentence stems offers solutions to linguistic theory’s apparent inability to account for how humans are able to “produce fluent multi-clause utterances” which “[exceed] human capacities for encoding novel speech in advance” (191) and for how native speakers are able to select native-sounding expressions from a range of possible, grammatically correct options.

Picking up on their ideas in his article “Improvisation, Creativity and Formulaic

Language,” Ian Mackenzie compares oral poetry and musical improvisation with the spontaneous generation of everyday speech, suggesting that similar processes are involved in all three areas. Just as Parry and Lord drew attention to the pervasive presence of formulas in oral poetry, Mackenzie writes that when we speak, “we routinely rely on a vast store of fixed, prepatterned phrases, which we use more often than we generate locutions entirely from scratch” (2000, 173). He critiques the Chomskyan idea that we generate sentences by taking individual words from our vocabulary and deploying them according to a set of learned grammatical rules; instead he suggests that our speech is primarily composed of memorized,

11 Formulas and the Building Blocks of Ṭhumrī Style pre-fabricated chunks of material, which we piece together in acts of communication. The basic units of language, he suggests, are not words, but groups of words. He cites examples of these: they include the idiomatic phrases “in a nutshell,” “by the way,” “for that matter,”

“what on earth” and “beside the point,” as well as the expression “how do you do?” (174).

Mackenzie suggests that we use these pre-fabricated chunks of material with flexibility. He describes them as “semi-fixed” and points out our capacity to create variations of familiar phrases and to combine and recombine learned expressions in imaginative ways (175).

Alison Wray has taken a detailed, multi-faceted approach to the analysis of formulaic language in her books Formulaic Language and the Lexicon (2002) and Formulaic

Language: Pushing the Boundaries (2008). She has looked, for example, at formulaic language in child and adult language learning, its possible causes and origins, its role in humor and its relationship with thought, as well as many other issues. She is interested both in psychological and social reasons why humans use formulaic language. Her arguments rest on the theory that “the mental lexicon is heteromorphic”; that is, that “linguistic material is stored in bundles of different sizes,” such that “the mental lexicon contains not only morphemes and words but also many multiword strings, including some that are partly lexicalized frames with slots for variable material” (2008, 12). She labels these linguistic units “Morpheme Equivalent Units” (MEUs) and suggests that these form the basis for the formulas that appear in spoken language.

Wray defines a MEU as “a word or word string ... that is processed like a morpheme, that is, without recourse to any form-meaning matching of any sub-parts it may have” (2008, 12).

Amongst her examples of multiword strings are “‘true’ idioms—a set of…particularly evocative multiword strings that express an idea metaphorically, such as ‘kick the bucket,’

‘red herring,’ and ‘raining cats and dogs’”; “expressions that are metaphorical but, arguably, are less distant from the literal meaning than idioms, such as ‘set one’s store by,’ ‘take stock,’

12 Analytical Approaches To World Music 2.1 (2012)

‘watch one’s back’”; “collocational associations, such as, in academic prose, ‘fully developed’ and ‘highly complex’” (2008, 10); and “partly-fixed frames,” which “have some fixed material along with slots that permit variation for other material,” including the frame

“NP be-TENSE as ADJi as ADJi can be,” which can be realized variously as “the elephant was as big as big can be,” “a flea is as small as small can be,” and “that tortoise will be as slow as slow can be” (2008, 16).

Wray proposes that that the way in which linguistic units enter the mental lexicon is determined by a default strategy which she labels “Needs Only Analysis” (NOA). According to this strategy, linguistic material enters the lexicon of the native speaker as it is heard, and is analyzed into smaller chunks only in response to particular, real-world needs. Wray writes,

“The process of analysis which the [native speaker] child engages in [is] not that of breaking down as much linguistic material as possible into its smallest components. Rather, nothing

[is] broken down unless there [is] a specific reason” (2002, 130, quoted in Wray 2008, 17).

Accounting for this psychologically, she argues that there is a “premium on working memory capacity,” with the result that “it is advantageous during linguistic processing to avoid unnecessary real-time encoding and decoding” (2008, 69). The use of already-encoded MEUs

(often multiword strings), then, minimizes the processing required in order to speak, with the result that, of the range of possible grammatically correct ways of expressing an idea, only certain material becomes “privileged as idiomatic” (2008, 18).

Wray also offers a social explanation for what she calls “formulaicity” in language, writing that the NOA strategy offers “a linguistic solution to a non-linguistic problem, namely, the promotion of self” (2008, 69). She attributes the motivation for using MEUs partially to “the social pressure to speak like others, something that can be achieved by adopting the multiword patterns already in use in the speech community” (18). She also notes that the social and psychological explanations she offers complement each other, since

13 Formulas and the Building Blocks of Ṭhumrī Style

“MEUs are not only easier for the speaker to encode but also easier for a hearer with similar knowledge to decode,” allowing a speaker to “select MEUs as a means of directing the hearer to certain actions, beliefs, and feelings, ... controlling how hearers perceive the speaker”

(2008, 69).

Wray offers a more sophisticated, nuanced explanation of the link between formulas and orality than that proposed by Parry, Lord and Ong. Rather than theorizing formulas as a direct consequence of orality, she suggests a number of different factors which predict greater or lesser formulaicity in different types of language. While she suggests that a “written medium” and a “complex society” are indicators of low levels of formulaicity, high formulaicity is determined by combination of the following five factors: “(1) the appropriacy of a formula in expressing exactly the desired message,” meaning that a formula will be more likely to be used if “a novel [i.e. not formulaic] formulation cannot capture all aspects of the meaning of the formula”; “(2) the estimated likelihood of the hearer understanding the formula”; “(3) the desire on the part of the speaker to signal identity through language,” such that high formulaicity is more likely to occur if the “producer wants to express insider identity even if receivers are excluded”; “(4) local conditions affecting the processing demands on the speaker and hearer,” formulas being more likely, for example, if the “producer is under processing pressure”; and “(5) the specific desire to express the idea in a novel way,” formulas being more likely in those situations in which “the producer does not desire to be specifically novel in expression” (2008, 57). She spells out the link between formulas and oral media of communication as an indirect one, such that “the medium of expression facilitates, rather than determines, the differences between text types” (57–58).

Wray’s approach has a number of advantages over those of her predecessors. It is holistic, bringing together both social and psychological factors to explain the presence of formulas in language. Also, by complicating any ideas of a direct, causal link between formulas and

14 Analytical Approaches To World Music 2.1 (2012) orality, Wray’s theories can account for the presence of formulas in written literature and notated musical traditions and for differences in the extent and type of formulas that occur in different domains. Furthermore, Wray raises the possibility that formulas serve a purpose not only for the people who use them, for example by ameliorating the problems caused by the limitations of human memory, but also for the people who hear them. She points out that formulas play a role in establishing social boundaries that might include both a speaker and a listener, and that when a speaker uses a formula also familiar to the listener, it renders language more easily intelligible by that listener (2008, 20–21).6

This article will suggest applying linguistic theories to music. The act of drawing comparisons between improvising and speaking is nothing new. Berkowitz, for example, explores connections between speech and musical improvisation at length, evoking a centuries-old tradition of using speech analogies to describe musical performance. He suggests that similar cognitive processes are involved in learning to improvise and learning to speak. He also explores other parallels between music and language, including in the way in which listeners experience them and in the cognitive processes involved during performance

(2010, 97–118, 145–152).

Likewise, I would like to suggest that the theories of formulas in language I highlighted above are highly applicable to the study of North Indian classical music. Pawley and Syder, along with Wray, point to the role that formulas play in reducing the cognitive processing required for encoding speech at speed; their descriptions are echoed in Slawek’s assertion that formulas enable musicians overcome the difficulties inherent in “keeping it going” in

6 Widdess makes a similar point in his discussion of “compositional principles” in North Indian classical music. He notes that their importance is not only “related…to the performer’s need to recall memorized material and invent new material that is grammatical” but also “to the listener’s need to engage with, comprehend, and be stimulated by an auditory experience that ... happens in real time” (2011, 188). He draws attention to the ways in which repeated musical patterns can generate listeners’ expectations, allowing them to engage with a performance as it progresses over time.

15 Formulas and the Building Blocks of Ṭhumrī Style performance, that is, in continuing to generate large amounts of musical material on the spur of the moment (1998).

Comparisons with spoken language would seem to offer a particularly appropriate way of understanding ṭhumrī. When they talk about ṭhumrī, musicians often compare it with speech.

When I was learning to sing ṭhumrī, my teacher, Sunanda Sharma, would sometimes congratulate me on singing “naturally, like speaking.” Many singers insist that, in ṭhumrī, it is of the utmost importance to convey the lyrics clearly; they caution against singing any words out of order (something which is acceptable in other classical genres) in case the lyrics become unintelligible. In his book on ṭhumrī, Peter Manuel compares ṭhumrī with the classical genre khyāl noting that in khyāl, but not in ṭhumrī, “text delivery is often blurred, broken, and generally unintelligible,” the emphasis of the performance being on “pure, abstract, melodic improvisation.” Noting the far greater weight attached to conveying the words of a ṭhumrī performance, Manuel arranges the different genres of North Indian music on a spectrum, placing ṭhumrī somewhere in between the two antipodes of what he calls

“text-dominated” genres and “music-dominated” ones (1989, 32–33). Even the very label applied to an improvised passage in ṭhumrī (bol) derives from the Hindi verb meaning “to speak” (bolna).

The arguments of Mackenzie and Wray concerning the link between formulas and language acquisition are also relevant to ṭhumrī. They both highlight the existence of linguistic formulas in order to advocate usage-based theories of language acquisition, which suggest that we learn to speak fluently in our mother tongue by first learning and repeating multi-word chunks of linguistic material verbatim, before we learn how to break them down, modify them and make up grammatical, idiomatic phrases on our own. This process parallels the way musicians learn to sing ṭhumrī. As a first step towards learning to create our own phrases of bol banāo, my teacher, for example, teaches her pupils first to memorize exactly

16 Analytical Approaches To World Music 2.1 (2012) the musical phrases she has created, then to break down and re-combine their component parts. On one occasion, early on in my musical training, my teacher had invited a tabla player to attend my singing lesson, so I could practice singing some of the phrases I had memorized to the sound of rhythmic accompaniment, providing the meter. One of the phrases she had taught me seemed too short, now that the meter was made explicit by the tabla player: it did not fill enough of the metrical cycle, which meant that I had to leave a gap before singing the next section. Hearing this, my teacher told me to fill that gap with the end of a different phrase that I had recently memorized. After demonstrating how I might do this, she said, “this is improvisation.”

An in-depth, comparative study of formulas in music and language is beyond the scope of this study. However, in the analysis I present in the next section, I will consider various ways in which theories of formulas in speech can usefully be applied to the analysis of ṭhumrī style. In particular, I will draw on concepts and terminology from work by Wray and

Mackenzie in order to make sense of some of the features I identify in ṭhumrī.

Formulas and Schemas in Western Classical Music

In his book Music in the Galant Style (2008), Robert Gjerdingen identifies a large number of repeating musical patterns in eighteenth-century Western classical music. Although the music he studies is the product of a notated musical tradition, these musical patterns nevertheless have a great deal in common with the formulas of oral traditions. Gjerdingen uses schema theory to make sense of his observations and labels the patterns he finds not

“formulas,” but “schemata.”7

7 Gjerdingen uses the Greek plural “schemata” in preference to the Anglicized “schemas.” Throughout this study, however, I use “schemas” and “formulas” (not “schemata” and “formulae”) in line with current practice in work by cognitive psychologists on schema theory and work by Wray and others on formulaic language.

17 Formulas and the Building Blocks of Ṭhumrī Style

Gjerdingen explains their presence in part by noting the improvisatory nature of much galant music. He compares music with performance of the contemporary Commedia

Dell’Arte, suggesting that musical schemas are the musical corollaries of the stock expressions used in this improvised theatrical tradition (8–9). He draws attention to the improvisatory nature of eighteenth-century musical performance, such that “[the score] often provided only a bare notation of the sequence of schemata, with the graces, ornaments, and elegant variation left to the skilled performer” (9–10). He points out that many musicians of the time were capable of improvising pieces on the spur of the moment. He mentions one documented instance in which two musicians jointly and simultaneously improvised a performance: he attributes this to their ability to “[connect] a string of well-learned musical schemata to form a seemingly spontaneous and continuous musical performance” (10).

Learning these schemas was a standard part of a galant musician’s training: Gjerdingen notes that students of music commonly made use of zibaldoni, notebooks of “exercises and rules,” which “provided an important repository of stock musical business from which a young composer could later draw” (10). Even the act of composing galant music in a notated score often involved the kind of spur-of-the-moment musical thinking more normally associated with improvisation, as a result of the economic environment of eighteenth-century music production, in which composers had to produce large amounts of music at speed. In this context, he notes the “obvious advantages” for composers of having a stockpile of memorized musical schemas, from which to put together their compositions (51). Additionally,

Gjerdingen points to other, social factors to account for the widespread presence of schemas in galant music; he associates them with courtly etiquette, comparing codes of musical schemas with the highly codified sets of gestures and manners that were required of those who wished to participate in contemporary courtly society.

18 Analytical Approaches To World Music 2.1 (2012)

Gjerdingen’s study is grounded in musical analysis. Based on his examination of work by a variety of eighteenth-century composers, he highlights and labels some of the most common schemas of galant music. Throughout his work, he shows how these musical patterns play out in specific compositions. He offers a highly flexible model of the functioning of schemas in music, showing ways in which they can be broken up, recombined, varied or made to overlap in individual compositions.

In my analysis of ṭhumrī, I adopt elements of Gjerdingen’s methodology in order to examine the musical formulas that occur there. Like Gjerdingen, I attempt to identify and distinguish different types of musical patterns and consider how they are deployed in specific performances. I will show ways in which ṭhumrī performances, like galant compositions, are largely made up of recurring musical patterns, which interact at different structural levels to produce a complex patchwork of musical events.

PART III: MUSICAL PATTERNS AND THE ANALYSIS OF ṬHUMRĪ PERFORMANCES

My analysis of ṭhumrī performances by a number of musicians has revealed the widespread presence of a variety of different types of recurring musical material and patterns.

This section will explore ways of analyzing and categorizing these musical phenomena and will consider how they might be taken into account when analyzing ṭhumrī style more generally. My analysis is corpus-based: my observations are grounded in my transcription and analysis of a large number of ṭhumrī recordings by prominent singers from about the last fifty years. I have supplemented this with interviews with performing musicians and by taking lessons in ṭhumrī and khyāl with the vocalist Sunanda Sharma.

In her publications, Wray devotes significant space to a discussion of the problems inherent in trying to define “formulaic language” (see, e.g. 2008, 9–11). She notes that there

19 Formulas and the Building Blocks of Ṭhumrī Style are many different types of linguistic patterning that might be contenders for the label

“formulaic” and whose occurrence might be explained in a variety of ways. She takes the strategic decision to commence her study with as broad a definition of “formulaic” as possible (2008, 4). In my study of musical patterns in ṭhumrī, I adopt a similar approach, taking a broad view of what might eventually turn out to be separate, discrete phenomena in order to look at as many aspects of ṭhumrī’s internal structure as possible. I consider ways of categorizing the different types of musical patterning found in ṭhumrī performances. For purposes of this study, a formula will be defined as any musical pattern which occurs repeatedly in ṭhumrī performances.

Exact Repetition or Abstract Strategies? Categorizing Formulas in Ṭhumrī

As a first stage in this analytical project, I would like to consider ways of differentiating the different types of musical patterning found in ṭhumrī performances. I suggest that it might be helpful to think about recurring musical patterns as lying on a spectrum, varying according to the degree of flexibility with which they are repeated, such that the exact repetition of chunks of musical material occupies one end and the recurring use of abstract, generational musical strategies occupies the other. I have highlighted four points along that spectrum, so that the precise repetition of what I call “stock expressions” is seen as one extreme, the non- exact repetition of what I label “variable outlines” and “musical gestures” lies somewhere in the middle of the spectrum and recurrence of abstract musical “strategies” lies at the other extreme (see Figure 2).

For the purposes of this discussion, it is worth noting that this method of differentiating types of musical patterning is not based on the nature of the material itself, but on the way it is used. While some chunks of musical material might be repeated exactly, others might be

20 Analytical Approaches To World Music 2.1 (2012)

Figure 2. Different types of recurring musical patterns in ṭhumrī.

treated flexibly on repetition, and yet others else might reveal an underlying abstract strategy.

Furthermore, the same chunk of material might operate in all three ways.

Drawing on a cognitive theoretical model suggested by Jeff Pressing, Berkowitz draws a distinction between “musical ‘objects’ that must be committed to memory so that they can be produced spontaneously when improvising” and “musical ‘processes’ that must be learned and rehearsed so that they can be used to develop formulas in improvised performance”

(2010, 40). Berkowitz classes the formulas he refers to as examples of musical objects; meanwhile, he cites “transposition, variation and recombination” as examples of musical processes. Similarly, Wray distinguishes in language between the store of linguistic units that make up the mental lexicon and the knowledge of “the principles by which they combine”

(2008, 89).

I would like to suggest that the equivalents in my model of Pressing’s musical objects are what I label “stock expressions,” while my “strategies” are examples of musical processes.

Some of the strategies I identify in ṭhumrī are in fact the same processes that Berkowitz identifies in Western classical music and jazz. However, unlike Berkowitz, I choose to assign the label “formula” to both objects and processes. My analysis of the types of musical patterns that recur in ṭhumrī suggests that, in this case, it is better to consider repeated musical objects and repeated musical processes as representing opposite poles of a spectrum than to suggest an absolute, qualitative distinction between these different types of musical

21 Formulas and the Building Blocks of Ṭhumrī Style patterns. In the analysis that follows, I will draw attention to types of recurring musical patterns that blur the boundary between object and process. I would like to suggest that, in the case of ṭhumrī at least, the distinction between musical objects and musical processes is not clear-cut.

Stock Expressions

On the far left hand of the spectrum in Figure 2, I have identified what I have called stock expressions: often very short chunks of material, these are repeated exactly or nearly exactly.

Figures 3 and 4 (Audio Examples 1 and 2) show two renditions of the word “bedardi” from a performance of the ṭhumrī “Bedardi balama” by Girija Devi (1993a): these provide a good example of the recycling of short segments of musical material within a single performance.

The principal difference between these two is the addition of upper Sa in the second example.8

Figure 3. Extract from Devi (1993a), ṭhumrī in rāg pīlū, 05:23 to 05:28.

Audio Example 1

Figure 4. Extract from Devi (1993a), ṭhumrī in rāg pīlū, 07:29 to 07:34.

Audio Example 2

8 See Appendix for an explanation of the notational conventions of North Indian classical music, covering both the use of letters to represent scale degrees and the use of Western staff notation.

22 Analytical Approaches To World Music 2.1 (2012)

Figures 5, 6, 7, 8 and 9 display another example of a stock expression; this time, it is in a performance by Rasoolan Bai (2007) and consists of a distinctive melodic pattern that always sets the end of the word “sā̃variyā.” (It is marked with brackets above the staff.)

Figures 10, 11, 12 and 13 show extracts from a ṭhumrī performance by Bhimsen Joshi

(2002a); these feature another stock expression. Here, the underlying schema consists of four

Figure 5. Extract from Rasoolan Bai (2007), ṭhumrī in rāg bhairavī, 07:07 to 07:14.

Audio Example 3

Figure 6. Extract from Rasoolan Bai (2007), ṭhumrī in rāg bhairavī, 07:59 to 08:03.

Audio Example 4

Figure 7. Extract from Rasoolan Bai (2007), ṭhumrī in rāg bhairavī, 08:06 to 08:11.

Audio Example 5

23 Formulas and the Building Blocks of Ṭhumrī Style

Figure 8. Extract from Rasoolan Bai (2007), ṭhumrī in rāg bhairavī, 08:59 to 09:06.

Audio Example 6

Figure 9. Extract from Rasoolan Bai (2007), ṭhumrī in rāg bhairavī, 09:10 to 09:14.

Audio Example 7

Figure 10. Extract from Joshi (2002a), ṭhumrī in rāg tilaṅg, 06:00 to 06:09.

Audio Example 8

Figure 11. Extract from Joshi (2002a), ṭhumrī in rāg tilaṅg, 06:56 to 07:02.

Audio Example 9

24 Analytical Approaches To World Music 2.1 (2012)

Figure 12. Extract from Joshi (2002a), ṭhumrī in rāg tilaṅg, 08:43 to 08:51.

Audio Example 10

Figure 13. Extract from Joshi (2002a), ṭhumrī in rāg tilaṅg, 14:08 to 14:12.

Audio Example 11

Audio Example 12

elements: 1) an optional lead-in, including the pitches N and/or S˙; 2) a pause on S˙; 3) an oscillating figure, using the pitches N and S˙; and 4) a figure tracing the outline

NPMGMPS˙NS˙. Note in these examples that the melodic material remains nearly exactly the same on each rendition, while the lyrics change. Joshi also sings another rendition of this formula in the un-texted ālāp section (see Audio Example 12, 03:05 to 03:13).

Girija Devi also re-uses musical material in two separate performances of the same composition: this can be seen in Figures 14 and 15, which show extracts from two different performances of the ṭhumrī “Kaise likhū̃ patiyā” (Devi 2001a and Devi 2002). Figures 16, 17,

18, and 19 show a similar figure; this time, they occur in her performance of a different composition, also in rāg pīlū (Devi 1993a).

My analysis has revealed an enormous number of instances, like these, in which short chunks of melodic material crop up repeatedly in ṭhumrī performances. Sometimes, as in the first example, this happens within a single performance; sometimes, in two performances of

25 Formulas and the Building Blocks of Ṭhumrī Style the same composition by the same singer; and sometimes, in performances that appear to have little else in common. I would like to suggest that these are akin to the memorized

Figure 14. Extract from Devi (2001a), ṭhumrī in rāg pīlū, 04:22 to 04:30.

Audio Example 13

Figure 15. Extract from Devi (2002), ṭhumrī in rāg pīlū, 05:44 to 05:54.

Audio Example 14

Figure 16. Extract from Devi (1993a), ṭhumrī in rāg pīlū, 02:38 to 02:53.

Audio Example 15

Figure 17. Extract from Devi (1993a), ṭhumrī in rāg pīlū, 02:55 to 03:02.

Audio Example 16

26 Analytical Approaches To World Music 2.1 (2012)

Figure 18. Extract from Devi (1993a), ṭhumrī in rāg pīlū, 03:15 to 03:22.

Audio Example 17

Figure 19. Extract from Devi (1993a), ṭhumrī in rāg pīlū, 03:56 to 04:08.

Audio Example 18

“licks” of jazz performance, the stock phrases of epic poetry and the multi-word chunks that

Mackenzie identifies in everyday speech. Already memorized and possibly also even rehearsed, musicians can utilize them at appropriate points in their ṭhumrī performances: they are some of the basic units of musicians’ vocabularies.

Variable melodic outlines

Instances of repetition in ṭhumrī do not always involve short, isolated chunks such as these; sometimes significantly longer passages are repeated, too. Note the striking similarity of the extracts in Audio Examples 19, 20 and 21, taken from a single ṭhumrī performance by

Purnima Chaudhuri (1996). (Audio Example 19 is transcribed in Figure 20, below.) The extract in Audio Example 19 comes from the start of the performance, from 00:32 to 00:47.

The extract in Audio Example 20 occurs much later, from 06:49 to 07:03, and the extract in

Audio Example 21 is from 07:35 to 07:46.

This almost precise repetition of a longer passage is unusual; while a number of singers frequently repeat shorter chunks of melodic material verbatim, longer passages are more normally repeated with greater flexibility than is evident in this example. Rather than the

27 Formulas and the Building Blocks of Ṭhumrī Style

Figure 20. Extract from Chaudhuri (1996), ṭhumrī in rāg khamāj.

Audio Example 19

Audio Example 20

Audio Example 21

precise recycling of musical material, these longer passages more typically involve recurring melodic outlines, realized differently on each occurrence. In their variability, they are akin to the “semi-lexicalised sentence stems” described by Pawley and Syder, Wray’s “partly-fixed frames” (2008, 16) and Mackenzie’s “semi-fixed” expressions.

One example of a melodic outline, varied on each rendition, occurs in the performance of the ṭhumrī “Rasa ke bhare tore naina” by Rasoolan Bai (2007). Figure 21 shows two extracts from this performance. The first extract is transcribed on the top line and the second is beneath it. Note that a significant amount of melodic material is shared in these two extracts.

See also the opening of two different performances in rāg bhairavī by Girija Devi (Figure

22, Devi 2001b and Devi and Gurtu 1997). There are marked similarities in the ways in which she sings the first three phrases of these two different ṭhumrīs. Figure 22 serves as a transcription of the opening phrases of both these performances. For clarity, each successive phrase is transcribed on each successive line here. To see what she does in “Bājūbanda khula khula jāya” (Devi 2001b, Audio Example 24), read the notes in the rectangles (on the first and third lines), but ignore the circled notes (on the middle line). To see what she does in

“Bābula morā” (Devi and Gurtu 1997, Audio Example 25), read the circled notes, but ignore

28 Analytical Approaches To World Music 2.1 (2012)

Figure 21. Two extracts from Rasoolan Bai (2007), ṭhumrī in rāg bhairavī, 00:23 to 00:39 and 00:53 to 01:06.

Audio Example 22

Audio Example 23

Figure 22. Extract from Devi (2001b) and (Devi and Gurtu 1997), ṭhumrīs in rāg bhairavī.

Audio Example 24

Audio Example 25

Note that in these extracts, both performances employ a common melodic outline, despite the

notes in the rectangles. All the notes not enclosed in circles or rectangles are common to both performances. I have not transcribed exactly what happens at the start of the bottom line, as this differs too much between the two performances; however, on both occasions Devi uses this phrase to introduce śuddh (natural) Re, transcribed here in brackets as D natural. Note

29 Formulas and the Building Blocks of Ṭhumrī Style that in these extracts, both performances employ a common melodic outline, despite sometimes differing in the notes they use.

In the previous two examples, the musical material was not repeated exactly, but was treated with considerable flexibility. Girija Devi, in particular, displays great ingenuity in the way in which she varies melodic material on repetition. Note in this example the parallel between the ends of the first and last phrases in her rendition of “Bājūbanda khula khula jāya.” Towards the end of the first phrase (at the end of the top line here), she rests on Pa.

The phrase could have ended here, as it does in her performance of “Bābula morā.”

However, she chooses to extend it with the figure MG-P. Then, towards the end of the third phrase (on the bottom line), a similar thing happens. Here, she rests on Sa. Again, the phrase could have ended there as in “Bābula morā,” but instead Devi extends it with the figure ND-

S. This parallelism creates a satisfying musical echo and is typical of the subtle craftsmanship of Devi’s performances. It is worth noting that this example of what I call a “variable melodic outline” itself contains a stock expression in the pattern DPMPMGMG at the end of the second line, which is highly similar in the two performances and also occurs frequently elsewhere in

ṭhumrīs in bhairavī: this musical passage is indicative of the ways in which multiple different types of formulas might overlap and be nested within each other when they are realized in performance.

Many examples of the types of formulas I have discussed so far (what I have called

“stock expressions” and “variable melodic outlines”) are specifically associated with particular rāgs; in some cases, they serve a crucial function in signaling to the audience which of a number of similar rāgs is being performed. Many musicians emphasize the importance of repeated musical formulas in characterizing rāgs in khyāl, too, often defining rāgs with reference to certain “key phrases” (Bor 1999, 2). In his discussion of rāg in ṭhumrī,

Peter Manuel suggests that semi-classical rāgs are often more formulaic than classical ones,

30 Analytical Approaches To World Music 2.1 (2012) drawing attention in particular to the high formulaicity of rāg bhairavī and highlighting some of characteristic melodic patterns of a number of other prominent ṭhumrī rāgs (Manuel 1989,

195–211).

While noting the association between particular formulas and particular rāgs, Manuel also draws attention to remarkable similarities between some of the characteristic phrases of different rāgs, highlighting examples in which particular melodic outlines appear in more than one rāg, but transposed, so as to make use of different scales degrees in each different rāg. In one example, a common phrase starts on Pa when performed in rāg khamāj, on Sa when performed in rāg kāfī and on komal Ni when performed in rāg bhairavī. Manuel suggests that this phenomenon is a function of the nature of human memory, which may cause phrases to be shared “among different rāgas with certain structural affinities” (212–

221).

Gestures

Moving further to right on the spectrum shown in Figure 2, my survey of ṭhumrī style has revealed another type of recurring pattern, which I have called “musical gestures.” These are pitch-contour patterns that recur in ṭhumrī performances: they consist of a repeated overall melodic shape, even while the precise scale degrees and intervals through which they are realized might vary. One example of such a gesture may be found at the start of phrases sung by Girija Devi. Devi often opens a phrase or a section of improvisation with a distinctive, highly ornamented rising figure, which I have shown graphically as an upwards squiggle

(Figure 23). Examples of these can be found in the two bhairavī extracts transcribed in Figure

22 above, where they traced the outline NSGMPDP; these figures occur frequently in almost all of Devi’s ṭhumrī performances, regardless of the rāg or composition she is singing. Audio

Examples 26 to 33 are further instances of this. Audio Examples 26 to 30 are taken from her

31 Formulas and the Building Blocks of Ṭhumrī Style

Figure 23. Graphic representation of a rising figure in ṭhumrīs by Girija Devi.

Audio Example 26 Audio Example 30

Audio Example 27 Audio Example 31

Audio Example 28 Audio Example 32

Audio Example 29 Audio Example 33

performance of “Itanī araja morī māna” (Devi 1989), Audio Example 31 from “Hamase najariyā” (Devi 2004) and Audio Examples 32 and 33 from “Binati karat hū̃” (Devi 2001c).

Another musical gesture informs the extracts in Figures 24, 25, 26, 27, 28 and 29. I have labeled this gesture “downward chromatic sliding.” It is extremely common in ṭhumrī and is characteristic of the genre. In this gesture, musicians exploit the possibility, unique in semi- classical renderings of rāg, of being able to sing two versions of the same scale degree by moving chromatically down through them.

Variable melodic outlines and musical gestures are examples of formulas which blur the boundary between object and process, incorporating elements of both. As schematic musical outlines, involving the use of particular pitches in a particular order, they constitute musical objects in their own right. They also, however, constitute musical processes that can be applied to already existing formulaic material. Many of Devi’s opening upward gestures, for

32 Analytical Approaches To World Music 2.1 (2012)

Figure 24. Extract from Joshi (2002a), ṭhumrī in rāg tilaṅg, 08:39 to 08:42.

Audio Example 34

Figure 25. Extract from Badi Moti Bai (n.d.), ṭhumrī in rāg bhairavī, 03:34 to 03:41.

Audio Example 35

Figure 26. Extract from Sharma (2002), ṭhumrī in rāg miśra des, 11:15 to 11:22.

Audio Example 36

Figure 27. Extract from Devi (2002), ṭhumrī in rāg pīlū, 07:34 to 07:46.

Audio Example 37

Figure 28. Extract from Devi (2001a), ṭhumrī in rāg pīlū, 05:44 to 05:41.

Audio Example 38

33 Formulas and the Building Blocks of Ṭhumrī Style

Figure 29. Extract from Devi (1993b), ṭhumrī in rāg bhairavī, 04:41 to 04:44.

Audio Example 39

example, incorporate key formulaic patterns of the rāg that she is presenting. Note also in

Figure 22 that the melodic outline functions as a process that transforms MGP at the end of the first line into the parallel figure NDS at the end of the third.

Strategies

Moving to the right-hand end of the spectrum are what Widdess and Nooshin label

“compositional strategies” (2006) or what Steven Slawek calls “dynamic, generative programs” (1998). These do not imply any particular collections of pitches, or even specific pitch-contours, but rather are abstract principles which musicians may use in generating improvised passages. One such formula in ṭhumrī is what I have labeled the “end-rhyme strategy.” According to this strategy, a singer sings a number of successive phrases, all of which end with the same musical material. In passages that employ this strategy, other organizing schemas often also come into play. For example, each successive phrase often increases in complexity and/or utilizes a greater range than its predecessor.

Note the three successive ways in which Girija Devi sings the text “avanaki” in the

ṭhumrī “Kaise likhū̃ patiyā” (Figure 30, Audio Example 40). The first is relatively simple:

DNDP. The next two renditions become increasingly complex and introduce ever-higher pitches; each time, however, they retain the distinctive NDP at the end.

The end-rhyme strategy may be viewed as part of a broader family of strategies which I

have labeled “successive variation” strategies. According to such strategies, musicians make

34 Analytical Approaches To World Music 2.1 (2012)

Figure 30. Extract from Devi (2002), ṭhumrī in rāg pīlū.

Audio Example 40

a musical point of focusing on a relatively small amount of musical material, singing successive musical units which retain some aspects of the melody while varying others. Thus in the context of the end-rhyme strategy, the end of the phrase is kept the same, while the beginning is varied.

Another example of successive variation can be found in the ṭhumrī “Sajan tum kahe ko neha lagaye,” sung by Abdul Karim Khan (Figure 31). The top line in Figure 31 shows a short phrase, setting one line of text. After singing this, Khan then sings a number of shorter figures, each borrowing musical material from the initial phrase, but varying it on repetition.

Here, I have aligned the chunks vertically so as to show from which part of the opening line all the subsequent figures are derived.

Figure 32 shows another instance of successive variation. Here, there are three melodic characteristics that are shared in these phrases; I have indicated them with circles, rectangles and dotted rectangles. In all three phrase-units, Begum Akhtar sings the arpeggiated figure

SGP, circled here. In the first and third phrase-units, this is followed by the figure PDPMG, which I have marked with a solid rectangle. In the second and third phrase-units, she starts to sing the word “jā” to the pitches GRS, marked here with a dotted rectangle, in a figure which echoes the rapid PMG of the end of the first phrase-unit.

Another example of successive variation is shown in Figure 33. Here, each successive

phrase-unit explores different ways of singing “jāvo” that ascend melismatically to upper Sa

on “jā-” and conclude with a syllabic setting of “-vo” at upper Sa. Many melodic features are

35 Formulas and the Building Blocks of Ṭhumrī Style

Figure 31. Extract from Khan (2003), ṭhumrī in rāg tilaṅg.

Audio Example 41

Figure 32. Extract from Akhtar (2005), ṭhumrī in rāg miśra khamāj, 03:06 to 03:23.

Audio Example 42

36 Analytical Approaches To World Music 2.1 (2012)

Figure 33. Extract from Devi (1960), ṭhumrī in rāg khamāj, 09:53 to 10:10.

Audio Example 43

shared between these different settings. These include the figure GMP in phrase-units 1, 2, 3,

5, and 7; MPSNS in 2, 3 and 5, preceded by GMPN in 3 and 5; and N alternating with grace- note P in 6 and 7.

Another subset of successive variation strategies is the group of what I have labeled

“transposition strategies.” In a transposition strategy, a short musical figure is repeated, but transposed, so as to appear at different pitches, often successively higher. When this occurs, any distinctive rhythmic features are normally retained, as are the lyrics. Figures 34, 35 and

36 show examples of this in three different ṭhumrī performances: the transposed figures are labeled x, y and z in each respective example.

Successive variation strategies are very common in ṭhumrī; they would seem to be particularly appropriate to the aesthetic of bol banāo text elaboration. One of the defining features of ṭhumrī is the process of bol banāo, by which singers sing one piece of text to different melodic settings, each designed to bring out different emotional nuances in the words. In such passages, the text is kept the same while the melody is varied. (Figure 33 shows a very typical example of this.) The technique just discussed here of singing one melodic phrase in a number of slightly different but related ways, and sometimes to different texts, displays a similar inventive impulse and demonstrates a similar desire to explore the potential inherent in a limited amount of (musical and/or textual) material.

37 Formulas and the Building Blocks of Ṭhumrī Style

Figure 34. Extract from Devi (1993a), ṭhumrī in rāg pīlū.

Audio Example 44

x’’ x x'

Figure 35. Extract from Joshi (2002b), ṭhumrī in rāg kāfī.

Audio Example 45

y y’ y’’

Figure 36. Extract from Rasoolan Bai (1964), ṭhumrī in rāg jogiyā.

Audio Example 46 z z’ z’’ z’’’

Further Comments on Variation and Variety in North Indian Classical Music

Successive variation is a strategy that privileges the variation of musical material as a primary focus of musical interest within a short passage of music. It affords listeners the opportunity to marvel at the ingenuity of a musician who is able to create seemingly endless variants of small amount of musical material. This is not the only way in which variation occurs in North Indian classical music. Sometimes musical material is repeated, with variation, such that its different occurrences are separated by a period of some time, perhaps over the course of a single performance or in different performances. This article has already

38 Analytical Approaches To World Music 2.1 (2012) drawn attention to instances, for example, where musicians sing different variants of a common melodic outline, separated by a few minutes in performance, or even in different performances.

It seems likely that some variation occurs unconsciously, as a consequence of the limitations of human memory. In my singing lessons, my teacher told me to memorize and repeat large amounts of musical material. Often, my inability to memorize this material exactly would result in my varying it over the course of the lesson. (I would discover this later, when I got home and listened back to the recordings I had made of my singing lessons.)

However, in North Indian classical music, the variation of musical material on each repetition is not always merely a haphazard consequence of the nature of memory, but rather something for which musicians consciously strive. Musicians often vary even the composition itself

(including the mukhra), though they consider it “fixed.” My teacher has remarked that this is

“to avoid boredom,” both of the listener and the performer. In my lessons, she would sometimes demonstrate numerous ways of varying and ornamenting the compositions she taught me, singing different versions, one after another. This would normally continue for a few minutes, in an extended show of inventiveness and creativity. Sharma would then instruct me to listen to my recordings of my lessons, choose some five or six versions of the composition from the many that she had sung and memorize them, so as to be able to use them in my own performances. (Sharma’s demonstrations of variants of the composition constituted performances in themselves, informed by successive variation strategy.)

Sharma’s attitude is indicative of a more widespread premium on variety achieved through variation in North Indian classical music. There are social reasons for this. Napier

(2006) has written of the importance of “subtle variation” for North Indian classical musicians. He sees this as a result of two simultaneous and contradictory impulses that act upon them: for their performances to be considered praiseworthy, they are expected both to

39 Formulas and the Building Blocks of Ṭhumrī Style remain faithful to the style of their teachers and gharanas (pedagogical lineages) and also to demonstrate musical originality. Variation (rather than innovation) allows musicians to strike a balance between novelty and conservatism by singing in a way that recognizably adheres to the styles of their predecessors (and also to the rules of rāg grammar), without replicating the performances of other musicians.

In Wray’s account of formulaicity in language, she notes that the extent to which formulas appear in a particular context is partly determined by the extent to which “the producer [desires] to be specifically novel in expression” (2008, 57). In the case of ṭhumrī, a premium on variety mitigates against precise repetition but not against formulaicity per se.

Acknowledging this can help to account for the frequent occurrence of variation in the repetition of formulaic patterns in ṭhumrī. This is another aspect of ṭhumrī that is well suited to Wray’s holistic model of formulaicity. By avoiding drawing any direct link between formulas and orality and instead looking at the operation of a variety of social and psychological factors, this model can account not only for why formulas occur but also for why they occur differently in different contexts.

The Beginning-Middle-End Paradigm: Formulas and the Delineation of Phrase Structure

In his analysis of eighteenth-century stock musical schemas, Gjerdingen notes that certain patterns served particular roles in delineating the structure of phrases. He describes the schema he calls the Romanesca, for example, as an “opening gambit”; the Prinner, meanwhile, functions as a “riposte,” usually following an opening gambit such as the

Romanesca. Kofi Agawu has also written of the syntactic function of different types of musical material in the eighteenth century, detailing what he calls the

“beginning/middle/end” paradigm (1991, chapter 3). He notes that certain musical units have

“beginning” function and tend to appear at the start of phrases, others have “middle” function

40 Analytical Approaches To World Music 2.1 (2012) and tend to appear in the middle of phrases and yet others have “end” function and function as closing figures. In his discussion of the stock patterns of jazz, Berliner, too, notes that certain figures are more suited to the beginnings of phrases than others. He describes the

“catalytic” function of certain types of musical material, which may cause musicians to use it at the outset of solos (1994, 227).

In ṭhumrī, likewise, certain formulas seem to occur most frequently at particular points within phrases. I have already mentioned an example of this when I described the upward gestures that Girija Devi uses to commence new phrases. Other formulas tend to occur at the close of phrases and phrase-units. In her “Mitva mane nare,” for example, Begum Akhtar frequently closes phrase-units with the figure GRSNS (sometimes omitting the grace-note S).

Figures 37, 38, 39 and 40 show some examples of this. In this performance, this figure never occurs in the middle of a phrase; in all instances, Akhtar stops singing afterwards and takes a breath. Note that it sometimes occurs in successive units (see Figures 38 and 39): here, it functions both as a more universal stock expression and also as part of a local end-rhyme strategy. Note also in Figure 38 that the second rendition of the figure here is interrupted. It breaks on Ni, denying the listener the resolution that would have been provided by the final

Sa, the tonic. This increases the sense of closure when the figure is then repeated at the end of the next phrase-unit. (The closure created here is sufficient for Akhtar then to cease singing for two full vibhāgs, shown here as measures, before introducing a new musical idea.)

Figure 37. Extract from Akhtar (1990), ṭhumrī in rāg kāfī, 01:23 to 01:30.

Audio Example 47

41 Formulas and the Building Blocks of Ṭhumrī Style

Figure 38. Extract from Akhtar (1990), ṭhumrī in rāg kāfī, 02:00 to 02:17.

Audio Example 48

Figure 39. Extract from Akhtar (1990), ṭhumrī in rāg kāfī, 04:17 to 04:26.

Audio Example 49

Figure 40. Extract from Akhtar (1990), ṭhumrī in rāg kāfī, 06:35 to 06:41.

Audio Example 50

The repeated use of particular musical units in a particular position within phrases creates a sense of musical syntax; it allows formulas to serve as familiar landmarks for listeners, delineating the structure even of unfamiliar musical phrases. This also affords the opportunity for musicians to manipulate listeners’ expectations, for example by extending a phrase after a conventional closing figure or by interrupting closing figures so as to create a greater sense of resolution when they are finally performed in full (as in Figure 38). There are many other ways in which musicians exploit the distribution of formulas in performance in order to

42 Analytical Approaches To World Music 2.1 (2012) create and manipulate listeners’ expectations: musicians may, for example, repeatedly perform two formulas in succession, creating a sense of surprise when the first is not followed by the second. There is, however, no space to discuss this here.

CONCLUSION: FORMULAS AND THE ANALYSIS OF ṬHUMRĪ STYLES

This article has considered ways of analyzing the musical patterns that recur in ṭhumrī performances. It has introduced terminology for different types of formulas and considered how studies of formulas in other fields can shed light on ṭhumrī. Along with some theoretical groundwork, I have presented a partial analysis of formulas in ṭhumrī. I believe that the phenomenon of recurring formulas in North Indian classical music holds enormous potential for further investigation. In particular, future research could explore the extent to which formulas play a role in characterizing different musical styles.

Discussion of the differences between “semi-classical” and “classical” genres in North

Indian classical music touches upon a variety of parameters. There are differences in the social meaning of genres, in popular understandings of their history, in the way in which musicians introduce them on stage and in the way in which musicians and audiences interpret their lyrics. There are also musical characteristics that differ in classical and semi-classical genres, including the rāgs permitted and way in which musicians use ornamentation. In this article, I have drawn attention to what I call “successive variation” and “downward chromatic sliding”: both formulas are characteristic of ṭhumrī. I suggested that successive variation contributes to the production in ṭhumrī of a specifically semi-classical “bol banāo aesthetic.”

Downward chromatic sliding, meanwhile, is a formula that is particularly suited to semi- classical genres, since it requires that the singer be able to sing two versions of the same scale degree in close proximity, something only normally permitted in semi-classical performances. I would like to suggest that recurring musical formulas also play a role in

43 Formulas and the Building Blocks of Ṭhumrī Style delineating the boundary between classical and semi-classical genres. Since most ṭhumrī singers are also khyāl singers, further comparison between the use of formulas in ṭhumrī and in khyāl would be valuable.

Future research on formulas could also focus on the analysis of the individual style of particular musicians or groups of musicians. In her study of formulaic language, Wray discusses “the need to distinguish between something that is formulaic ‘in the language,’ so to speak, and something that is formulaic just for a particular individual or group” (2008, 11).

Likewise, Nicholas Magriel discusses the varying degrees of universality of patterns in North

Indian classical music in his article “The Barhat Tree” (1997). He distinguishes between a musical “idiolect,” which is a performer’s unique conception of the rāg of the performance, a

“dialect,” which might be shared by a number of musicians in the same pedagogical lineage, and a “language,” which constitutes the universal set of musical characteristics common to all renditions of a particular rāg (109). In ṭhumrī, there would seem to be multiple levels at which musical material might recur. At the broadest possible level, a particular pattern might be used commonly by a large number of ṭhumrī singers and might be considered ‘in the language’ of ṭhumrī. Another pattern might be shared primarily by singers of a particular gharana, or by singers who specialize in a particular ang (branch) of ṭhumrī style. Yet another pattern might characterize the style of one individual, or of a prominent teacher and his or her pupils. There would seem to be great potential in looking in detail at the formulas favored by particular musicians, analyzing not only their individual formulaic vocabularies but also their individual approaches to combining and varying formulas in performance. It would also be valuable to consider the extent to which formulas can be transmitted from teacher to pupil. Given the large amount of memorization involved in learning to sing, it would seem likely that pupils would commonly inherit formulaic turns of phrase from their

44 Analytical Approaches To World Music 2.1 (2012) teacher. These would seem to be rich areas for future study, promising to shed light on the nature of style in North Indian classical music.

REFERENCES

Agawu, V. Kofi. 1991. Playing with Signs: A Semiotic Interpretation of Classic Music. Princeton, N.J: Princeton University Press.

Berkowitz, Aaron. 2010. The Improvising Mind: Cognition and Creativity in the Musical Moment. New York: Oxford University Press, USA.

Berliner, Paul. 1994. Thinking in Jazz: The Infinite Art of Improvisation. Chicago: University of Chicago Press.

Bor, Joep. 1999. The Raga Guide: A Survey of 74 Hindustani Ragas. Wyastone Leys: Nimbus Records.

Gjerdingen, Robert O. 2007. Music in the Galant Style. USA: Oxford University Press, USA.

Lipiczky, Thom. 1985. “Tihai Formulas and the Fusion of ‘Composition’ and ‘Improvisation’ in North Indian Music.” The Musical Quarterly 71 (2): 157–171.

Lord, Albert Bates. 1965. The Singer of Tales. New York: Atheneum.

Mackenzie, Ian. 2000. “Improvisation, creativity, and formulaic language.” Journal of Aesthetics & Art Criticism 58 (2): 173–179.

Magriel, Nicolas. 1997. “The Barhat Tree.” Asian Music 28 (2): 109–133.

Manuel, Peter. 1989. Thumri in Historical and Stylistic Perspectives. Delhi: Motilal Banarasidass.

Meer, Wim van der. 1980. Hindustani Music in the 20th Century. The Hague: Martinus Nijhoff.

Napier, John. 2006. “Novelty that must be subtle: continuity, innovation and ‘improvisation’ in North Indian music.” Critical Studies in Improvisation 1 (3).

Nettl, Bruno. 1974. “Thoughts on Improvisation: A Comparative Approach.” The Musical Quarterly 60 (1): 1–19.

45 Formulas and the Building Blocks of Ṭhumrī Style

Nooshin, L. 2003. “Improvisation as’ Other’: Creativity, Knowledge and Power-The Case of Iranian Classical Music.” Journal of the Royal Musical Association 128 (2): 242–296.

Ong, Walter J. 1982. Orality and Literacy: The Technologizing of the Word. New York: Methuen.

Parry, M. 1987. The Making of Homeric Verse: The Collected Papers of Milman Parry. Oxford University Press, USA.

Pawley, A., and F. Syder. 1983. “Two Puzzles for Linguistic Theory: Nativelike Selection and Nativelike Fluency.” In Language and Communication., ed. J. C. Richards, and R. W. Schmidt. London: Longman.

Rubin, David C. 1995. Memory in Oral Traditions: The Cognitive Psychology of Epic, Ballads, and Counting-Out Rhymes. New York: Oxford University Press.

Schank, Roger C, and Robert P Abelson. 1977. Scripts, Plans, Goals, and Understanding: An Inquiry into Human Knowledge Structures. Hillsdale, N.J. : Lawrence Erlbaum.

Slawek, Stephen. 1998. “Keeping it Going: Terms, Practices, and Processes of Improvisation in Hindustani Instrumental Music.” In In the Course of Performance: Studies in the World of Musical Improvisation, 335–368.

Smith, Gregory Eugene. 1983. “Homer, Gregory, and Bill Evans?: the Theory of Formulaic Composition in the Context of Jazz Piano Improvisation.” Ph.D diss., Harvard University.

Snyder, Bob. 2000. Music and Memory: An Introduction. Cambridge, Mass: MIT Press.

Treitler, Leo. 1974. “Homer and Gregory: the Transmission of Epic Poetry and Plainchant.” The Musical Quarterly LX: 333–372.

Widdess, Richard. 2011. “Dynamics of Melodic Discourse in Indian Music: Budhaditya Mukherjee’s Alap in Rag Puriya Kalyan.” In Analytical and Cross-Cultural Studies in World Music, ed. Michael Tenzer and John Roeder, 187–224. New York: Oxford University Press.

Widdess, Richard, and Laudan Nooshin. 2006. “Improvisation in Indian and Persian music.” Journal of the Indian Musicological Society 36–37.

Wray, Alison. 2002. Formulaic Language and the Lexicon. Cambridge: Cambridge University Press.

———. 2008. Formulaic Language: Pushing the Boundaries. New York: Oxford University Press.

46 Analytical Approaches To World Music 2.1 (2012)

DISCOGRAPHY

Akhtar, Begum. 1990. “Mitwa mane nare.” Thumris, Sawan, Ghazals. Music India CDNF 078.

———. 2005. “Na ja balam pardes.” Begum Akhtar – Thumri, Kajri, Hori, Sawan, Dadra. Universal LTCDNF 175D.

Badi Moti Bai. n.d. “Nahak laaye gawanavaa.” Badi Motibai, volume 2. Sangeet Kendra CD.

Chaudhuri, Purnima. 1996. “Chabi dekhla ja.” The Lyrical Tradition of Thumri 1: Purnima Chaudhuri. Makar MAKCD020.

Devi, Girija. 1989. “Itni araj mori maan.” Golden Raaga Collection – Girija Devi – Live Performance at the Thumri Festival, Patna, 1989. Times Music TDIC 006.

———. 1993a. “Be dardi balam.” Girija Devi: Songs from Varanasi. Nimbus NI5315.

———. 1993b. “Nahak laye gawanwa.” Yaadgaar thumriyan. Saregama CDNF 150032.

———. 2001a. “Kaise likhun patiyaan.” Pratiksha: Girija Devi. Sagarika H-CD-0005.

———. 2001b. “Baju band khuli khuli jaye.” Girija Devi – Live in Benaras, November 4, 2001. Virgin 7243 812253 2 5.

———. 2001c. “Binati karat hoon.” Pratiksha: Girija Devi. Sagarika H-CD-0005.

———. 2002. “Kaise likhoon.” Khayal and thumri, vol.2. Venus VCDSP-1127.

———. 2004. “Raga Bihag.” Girija Devi – Vocal. India Archive Music CD 1076.

Devi, Girija and Shobha Gurtu. 1997. “Babul mora.” Girija Devi, Shobha Gurtu: Classical Vocal. Saregama CDNF 150191 DDD. Original sound recording by The Gramophone Company of India Ltd.

Devi, Siddheshwari. 1960. “Jao wahin tum shyam.” Siddheshwari Devi – All India Radio Archival Recording Vol. 1. Akashvani Archive AIRH.

Joshi, Bhimsen. 2002a. “Sajan tum kahe ko neha lagaye.” Thumri – Classical Vocal and Instrumental. Sagarika H-CD-0006.

———. 2002b. “Piya to manat nahin.” Rasrang: Shringar and Bhakti. Saregama CDNF 150460.

47 Formulas and the Building Blocks of Ṭhumrī Style

Khan, Abdul Karim. 2003. “Sajan tum kahe ko neha lagaye.” Golden Milestones: Ustad Abdul Karim Khan Sahib. Saregama CDNF 150481.

Rasoolan Bai. 1964. “Kaun gali gaye shyam.” Poorab-ki-thumri, Rasoolanbai, vol.2. Sangeet Kendra SK023.

———. Accessed 2007. “Rasa ke bhare tore naina.” Downloaded from homepage.mac.com/patrickmoutal/macmoutal/rag.html.

Sharma, Sunanda. 2002. “He ma kari badariya barase.” Sunanda Sharma: Live in Ayodhya, 2002. Virgin 7243 592539 2 3.

APPENDIX: NOTE ON TRANSCRIPTION

As is conventional in the transcription of Indian classical music when using Western staff notation, I have transposed the pitches so that the middle tonic is written as middle C.

Accidentals apply only to the notes that they are immediately beside. (They do not, for example, last until the end of the measure in which they appear.)

I have shown the metrical cycle, or tāl, using bar-lines and rehearsal letters. Each rehearsal letter indicates the position of the sum, the first beat of the metrical cycle. Each cycle then consists of a number of vibhags, which I have shown here as measures. Each vibhag contains a number of matras or beats, which I have shown as quarter notes.

In the Indian version of solfege, the scale degrees of Indian classical music are normally called Sa, Re, Ga, Ma, Pa, Dha and Ni, where Sa is the tonic, Re is the second scale degree etc. In transcription, scholars normally refer to each scale degree by its initial letter, so that Sa becomes S, Re becomes R etc. In their natural (śuddh) form, these are the scale degrees of the

Western major scale. However, these may also be flattened or sharpened, depending on the rāg. If they are flattened (komal), then they are underlined. For example a flat Ni is written N.

The only scale degree that can be sharpened is Ma (tivra Ma): this is written Ḿ.

48